jakowenko / double-take

Unified UI and API for processing and training images for facial recognition.
https://hub.docker.com/r/jakowenko/double-take
MIT License
1.19k stars 92 forks source link

[BUG] most faces are not matching because under 100x100 #329

Open timknowlden opened 11 months ago

timknowlden commented 11 months ago

Describe the bug Detector showing red for match because most box sizes are below 100x100

Hardware Unraid OS

Additional context Frigate, doubletake and deepstack

Is this a frigate or doubletake config issue, how can I increase the image size and box? 1000021953

DieMetRik commented 11 months ago

I have the same issue with deepstack detector. I switched to Compreface it solve issue with low % of face recognition

If you want to increase resolution of box -> increase resolution in camera settings and in Frigate detect section frigate config file:

    detect:
      width: 1920
      height: 1080
      fps: 5

You also can decrease value in double-take config

detect:
  match:
    # minimum area in pixels to consider a result a match
    min_area: 10000 

min_area it's size of orange box in pixel: 100 100 = 10000 if you need 80 80 = 6400